trading-manifold
Pass
Audited by Gen Agent Trust Hub on Feb 20, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- PROMPT_INJECTION (LOW): Potential for indirect prompt injection from untrusted external data. * Ingestion points: Market questions, descriptions, and metadata are retrieved from the Manifold API in functions such as search_markets and get_market (SKILL.md). * Boundary markers: None present; API data is interpolated directly into agent context. * Capability inventory: The skill possesses capabilities to place bets, sell shares, and create new markets, which could be misused if an agent follows instructions embedded in market descriptions. * Sanitization: No validation or escaping of the text fields returned by the external API is performed.
Audit Metadata